Calculating area of polygon per slope category

1552
5
08-04-2011 06:05 AM
SandyLyle
New Contributor
Hello,

I'm after opinions on the best way to achieve the following task. Opinions diverge as to whether raster or vector processing is most efficient for large datasets.

Input A:
vector dataset containing several slope categories (eg 1-6)

Input B:
vector land parcels or vector erosion zones; can contain over 500'000 features

Output required:
Area in m2 per slope category within each polygon in input B (output als statistics per polygon sufficient)

Other factors:
Performance should be optimized for handling large datasets/regions.
Software: ArcGIS 10, ArcInfo, incl. spatial analyst; FME

Questions:
- does someone have experience of this kind of processing?
- which tools re suited to the task/would you recommend?
- comparison of using vector or raster geoprocessing?

Any tips/hints welcome!

Thanks,
Sandy
0 Kudos
5 Replies
MarkEllis
New Contributor II
I haven't geoprocessed that many features since v9, but right away would lean toward using raster.  You might be looking at overnight runs to let the machine churn through the large number of records. 

A factor to consider is how precise your data must be.  Consider the number of raster cells you will need versus the number of polygons you'll process.  If you need 10 million grid cells, but only half a million polygons, the latter might work better. Are your slope values from a 30m DEM? Your grid cells should not be less than 10m.  If from LIDAR with cm accuracy, this would give you vastly different needs.

Raster processing would be simple, really.  Assign values 1-6 for Input A.  Add this (using the raster calculator) to values for Input B (values in multiples of 10) so that the possible output values are unique.  Every combination of A and B has a unique sum in the output.
0 Kudos
SandyLyle
New Contributor
Thanks for the inputs!
Slope is calculated on a 2m DEM.
Smallest island of any slope category is 10x10m.
Does B (vector input) need to be rasterized first?
0 Kudos
DaleHoneycutt
Regular Contributor
I'd go the most direct route using the Intersect tool followed by the Summary Statistics tool to calculate area -- the Statistics field would be SUM of shape_area and your Case fields would be both slope category and erosion zone.  No need to convert to raster.

Since you've got a large number of polygons, before running Intersect, turn off background processing and exit any other applications that chew up memory (like Firefox).
0 Kudos
SandyLyle
New Contributor
Hi Dale,

Thanks a lot for your inputs. Interesting that you go the vector route. Have the improvements in AGD10 made the difference? So far I have more suggestions for the raster route.

About to conduct some tests.

Sandy
0 Kudos
DaleHoneycutt
Regular Contributor
Vector overlay should work just fine, but upon re-reading your original post, performance is an issue and raster overlay will be faster than pure vector overlay.  But if this is just a one-off query, I'd still go vector.  But if you're building some sort of service that'll be used frequently, rasterizing your static data (parcels) would be an acceptable approach.  But of course you'll have to deal with resolution issues as others have posted.  I'm not a raster guy, but I'm thinking you'd just use the Zonal Statistics tool (after rasterizing your land parcels).
0 Kudos